ViTAdapterModelSize

Enums · AiDotNet.Enums

Defines the model size variants for ViT-Adapter.

For Beginners

ViT-Adapter adds spatial prior modules to plain Vision Transformers so they can handle dense prediction tasks like segmentation. The sizes correspond to different ViT backbone sizes (Small, Base, Large).

How It Works

Technical Details: Each size uses a different ViT backbone with adapter modules that inject multi-scale spatial priors. The adapter modules are lightweight and add only a small percentage of extra parameters on top of the base ViT.

Reference: Chen et al., “Vision Transformer Adapter for Dense Predictions”, ICLR 2023 Spotlight.

Fields

Field Summary
ViTAdapterModelSize Base ViT-Adapter-B: Base variant based on ViT-Base (86M params).
ViTAdapterModelSize Large ViT-Adapter-L: Large variant based on ViT-Large (304M params).
ViTAdapterModelSize Small ViT-Adapter-S: Small variant based on ViT-Small (48M params).
int value__